Infection with Mycobacterium tuberculosis afflicts millions of individuals worldwide with some estimates suggesting one third of the world’s population harbors latent tuberculosis infection.1 Primary infection in immunocompetent individuals sometimes results in containment but not eradication of the tuberculosis organism, which can remain dormant in the host for many decades. Latent tuberculosis infection is the name of this state of dormant infection, often defined as an asymptomatic state with the presence of tuberculosis organisms in lung tissues without any clinical or radiologic symptoms of an active disease.2 The importance of accurately diagnosing latent tuberculosis lies in the potential for reactivation to active and transmissible disease when the host is immunosuppressed by comorbid illness or medical therapy.
Reactivation is of particular relevance to patients with kidney failure receiving dialysis. Because kidney failure is associated with widespread immune system dysfunction,3 dialysis patients have a risk that is 10 to 25 times greater than the general population of reactivation to active tuberculosis disease.4 Reactivation poses substantial risk of transmission to other patients in hemodialysis units, many of whom are frail and at risk of adverse outcomes,5 as well as dialyzing in close proximity to each other. Moreover, in kidney transplantation patients, unrecognized latent tuberculosis can lead to fatal disseminated disease as a consequence of immunosuppressive therapy to prevent organ rejection. Accurate diagnosis is of tremendous importance in this population.6
There are several diagnostic tests available to determine the presence of latent tuberculosis infection: the tuberculin skin test (TST) or the interferon-γ release assays: QuantiFERON-TB Gold (QFT-G) and ELISPOT (T-SPOT.TB). These tests have been tested and validated extensively in the general population.7 The TST has historically been the most common and readily available test evaluated in detecting latent tuberculosis in immunosuppressed populations.4 Newer literature has suggested that the QFT-G or T-SPOT.TB may provide greater diagnostic utility in immunosuppressed populations, who may be anergic to the TST, and therefore more likely to have a false-negative TST result.4,8 The diagnostic performance of these tests remains poorly understood in patients with on hemodialysis. Further complicating the accurate detection of latent tuberculosis is the lack of a diagnostic gold standard. Popular standards include recent contact with someone who has been infected with tuberculosis, history of active disease, or a chest x-ray indicative of infection. Some studies also use a combination of these factors; however, there is no definitive agreement across the literature on the gold standard applicable in clinical practice.2
We hypothesized that the interferon-γ release assays will have greater diagnostic accuracy in determining latent tuberculosis infection in patients on hemodialysis when compared to the TST and against the variety of available gold standards (recent contact, history, and suggestive chest x-ray). To test this hypothesis, we conducted a systematic review of studies evaluating the diagnostic accuracy of the TST, QFT-G, and T-SPOT.TB in this population.
Study Selection and Characteristics
Our initial search strategy retrieved 2,039 unique citations for screening. Of these, 102 articles were selected for full-text review and 17 met the criteria for inclusion in the review. A flow diagram outlining the selection strategy can be seen in Figure 1.9 Thirteen studies evaluated the TST,4,8,10–20 10 studies evaluated QFT-G,8,11,15,16,18,20–24 and three studies evaluated the T-SPOT-TB test.4,18,22 Four studies considered populations in the United States,13,14,18,19 and four studies considered populations in Turkey.10,12,15,21 Two studies also represented each of Switzerland8,16 and South Korea.11,22 The remaining studies were from Canada,4 Japan23 Taiwan25 Spain20 and Belgium17 The characteristics of the selected studies are summarized in Table 1.
There was heterogeneity between the definitions used as a gold standard for latent tuberculosis infection across the selected studies. Thirteen studies had sufficient information to determine diagnostic accuracy compared to a gold standard of previous contact with someone infected with tuberculosis,4,10,12,13,15–21,25 eight studies demonstrated the diagnostic accuracy compared to a gold standard of a history of tuberculosis disease,4,10,12,15,17,21,23,25 and seven studies provided a gold standard of previous radiologic evidence.4,12,14–16 Five studies also compared to a gold standard that encompassed a combination of the previously mentioned risk factors.8,11,12,16,20,22
Tuberculin Skin Test
According to the fixed effects models, the TST had a pooled sensitivity of 31% (26%–36%, 95% confidence interval) and a pooled specificity of 63% (60%–65%) across eight studies with the most common latent tuberculosis proxy being previous contact (Fig. 2 and 3) and when the cutoff for a positive TST was defined as 10 mm induration. When considering the gold standard as a previous chest x-ray indicative of tuberculosis, the pooled sensitivity was 47% (35%–60%), and the pooled specificity was 63% (60%–67%); a gold standard of a previous tuberculosis contact had a pooled sensitivity of 25% (21%–29%), a pooled specificity of 64% (62%–67%), a history of tuberculosis disease had a pooled sensitivity of 62% (51%–72%), and a pooled specificity of 64% (61%–67%). A separate analysis performed on three studies and considering a positivity threshold of 5 mm determined that the pooled sensitivity was 19% (14%–23%) and the pooled specificity was 79% (70%–87%) (see Table S1, SDC,http://links.lww.com/TP/B67).
The QFT-G test had a pooled sensitivity of 53% (46%–60%, 95% confidence interval) and a pooled specificity of 69% (65%–72%) across nine studies with the most common latent tuberculosis proxies being previous contact or a combination of contact, chest x-ray, and history of tuberculosis (Fig. 2 and 3). When considering specific proxies, the pooled sensitivity and specificity for a chest x-ray indicative of previous tuberculosis infection were 59% (39%–76%) and 68% (61%–75%); for previous contact with someone infected with tuberculosis, the pooled sensitivity and specificity were 41% (33%–48%) and 62% (58%–67%); and for a history of tuberculosis disease, the pooled sensitivity and specificity were 70% (58%–82%) and 64% (60%–68%), respectively. Using a combination of proxies, wherein latent tuberculosis was defined as being one or multiple of the risk factors considered as a gold standard, the pooled sensitivity was 71% (62%–79%) and the pooled specificity was 82% (75%–87%) (see Table S2, SDC,http://links.lww.com/TP/B67).
The T-SPOT.TB test had a pooled sensitivity of 50% (42%–59%, 95% confidence interval) and a pooled specificity of 67% (61%–73%) across three studies with the primary gold standard being previous contact with someone infected with tuberculosis or a combination of risk factors (Fig. 2 and 3). Inclusion of indeterminate results as a negative test resulted in a pooled sensitivity and specificity of 35% (25%–46%) and 68% (62%–75%), respectively. Inclusion of indeterminate results as a positive test resulted in a pooled specificity and specificity of 38% (27%–48%) and 63% (56%–69%), respectively.
Measures of Consistency and Heterogeneity Across Studies
Heterogeneity across the evaluated studies was high for all three diagnostic tests. This heterogeneity may be a consequence of differences in interpreting gold standards across the considered studies. Among studies determining the sensitivity of the TST, QFT-G, and T-SPOT, the measures of inconsistency (I-squared) were 87.8%, 83.8%, and 92.1%, respectively, when applying a fixed effects model. Among studies determining the specificity of the TST, QFT-G, and T-SPOT, the measures of inconsistency were 97.2%, 92.5%, and 76.8%. All of these measurements were statistically significant at a P value less than 0.001 (Fig. 2 and 3). Rates of Bacillus Calmette-Guérin (BCG) vaccination were also heterogeneous across studies and ranged from 2.6% to 55%. The rate of BCG vaccination showed no correlation with the overall specificity or sensitivity (P = 0.39 between BCG rate and sensitivity, P = −0.166 between BCG rate and specificity).
To mitigate this high level of heterogeneity, a random effects model was also applied to determine the pooled estimates and confidence intervals for each of the diagnostic tests. The pooled point estimates were not significantly different between the fixed effects and the random effects models; however, the confidence intervals were larger for the QFT-G test and extremely wide when considering both the TST and the TSPOT (Figure S1, Figure S2, Table S3, Table S4, SDC,http://links.lww.com/TP/B67).
Reporting Quality and Risk of Potential Bias
Several sources of bias outlined in the QUADAS tool were common across the studies considered. Methods of patient selection were unclear in many studies. The length of time between the determination of the reference standard and the administration of the index test was uncertain. It was also unclear in many studies if the assessment of the index tests was blinded to the results of the reference standard. Some risk of bias was introduced because of the absence of a true clinical gold standard for diagnosis of latent tuberculosis infection. Although all studies used criteria commonly accepted as proxies for latent tuberculosis, histologic confirmation of dormant mycobacterium in lymph nodes or lungs as a gold standard is not feasible in a large observational study and was not done (Fig. 4).
In our systematic review and meta-analysis of studies evaluating screening strategies for latent tuberculosis infection in patients on hemodialysis, we found that the QFT-G or T-SPOT tests were more sensitive than using the TST across a variety of possible diagnostic standards while offering a comparable measure of specificity. These findings were consistent across various definitions of latent tuberculosis infection and when indeterminate results were considered positive or negative. These results suggest that the current clinical practice of using the TST for screening patients on hemodialysis for latent tuberculosis infection should be reexamined. Patients on dialysis are more likely to develop active tuberculosis and to suffer considerable morbidity and mortality from the disease. Furthermore, the presence of impaired cell mediated immunity and high false-negative TSTs are highly prevalent in this population.15,18 Sensitive and specific detection and treatment of latent tuberculosis in potential transplant recipients is of major importance because risk of reactivation rises substantially with induction and maintenance immunosuppression. Failure to accurately detect latent tuberculosis in these candidates can place patients at substantial risk of death or graft loss.6 In addition, patients can be prescribed prophylactic treatment for latent tuberculosis reflection which may lead to uncommon, although unnecessary, harm as a result of misdiagnosis.26
To our knowledge, our systematic review is the first to examine the diagnostic accuracy of the TST, QFT-G, and T-SPOT in patients on hemodialysis based on sensitivity and specificity. A previously published systematic review considered the prognostic value of several tests in dialysis and transplant patients based on reported odds ratios with respect to clinical risk factors for latent tuberculosis and had similar and consistent findings.27 Our study complements and extends the findings of this previous review by providing estimates of diagnostic accuracy (sensitivity or specificity) rather than odds ratios and is therefore more relevant for clinicians evaluating the accuracy of a diagnostic test.
Current clinical guidelines suggest using the TST for diagnosis of latent tuberculosis in patients on hemodialysis and other immunosuppressed populations, possibly based on estimates for cost-effectiveness derived from general population data28–30 These estimates may not be accurate in patients on hemodialysis for multiple reasons. Firstly, these patients are more likely anergic, and as a result, the diagnostic accuracy of the TST is diminished.4,8 Secondly, the risk of reactivation to active, transmissible disease in this population may be substantially higher, particularly after organ transplantation, suggesting that a lower sensitivity may have significant clinical consequences not observed in the general population.31 Finally, as interferon-γ release assays become more widely used in North America and worldwide, their cost is likely to decrease, particularly in the long term in accordance with increasing demand in the presence of price elasticity as a result of substitute products.32 In fact, a recent cost-effectiveness study has concluded that the use of interferon-γ release assays as opposed to the TST may be cost-effective in the hemodialysis population.33 Although the application of the QFT-G or TSPOT.TB tests may be more expensive on a per-test basis,33 the resulting benefits in preventing potential reactivation episodes afforded by the more expensive test may be an acceptable expense.
Our review had several strengths. The search strategy included multiple electronic databases in an attempt to ensure that all of the published literature examining the diagnostic accuracy of the TST, QFT-G, and T-SPOT were captured. Furthermore, we manually searched the bibliographies of included articles to ensure the sensitivity of our search strategy. We considered the quality of the included studies applying validated criteria34 to ascertain risk of bias and applied a random effects model to mitigate the heterogeneity across studies. Lastly, the total sample size across the 17 included studies totalled a sample of 2,899 patients with kidney failure on hemodialysis.
There were also several limitations present in the study. A study that performed serial testing in hemodialysis patients with interferon-γ release assays found that both conversion and reversion of TSPOT and QFT-G were associated with risk for previous latent tuberculosis infection, but that the interpretation of these rates must be further evaluated.22 Few studies are available on the diagnostic accuracy of the TSPOT test for those on hemodialysis, and more research is required with respect to this diagnostic test to increase the power in pooled TSPOT results. A major limitation of any study in this area is the lack of a true clinical gold standard for latent tuberculosis infection diagnosis and overall heterogeneity of the gold standards used across studies and between considered populations. Further heterogeneity was present when considering the definition of a close contact and there may be inaccuracies using a self-reported history of tuberculosis disease. Moreover, diagnosis of latent tuberculosis infection relies on historic or radiographic evidence of previous tuberculosis, and as such is a flawed standard. Finally, many of these studies evaluated testing in regions of the world having higher overall incidence and prevalence of tuberculosis than western countries, as well as different BCG vaccination rates, and therefore it may be difficult to generalize these results to all settings.35
In conclusion, our systematic review and meta-analysis found that the performance of the TST in patients with kidney failure on hemodialysis was poor. This calls into question the continued practice of using the TST to screen for latent tuberculosis infection in these patients. The QFT-G and the T-SPOT.TB tests are more sensitive than the TST with regards to the diagnosis of latent tuberculosis in the population of patients on hemodialysis and may result in more accurate screening results. Further research to strengthen the evidence concerning the cost-effectiveness of using interferon-γ release assays, QFT-G and TSPOT, in this population is needed.
MATERIALS AND METHODS
Data Sources and Searches
We retrieved information from the following databases in collaboration with a medical librarian (K.M.): PubMed, Scopus, EMBASE, and Cochrane Database of Systematic Reviews. Our search of these databases ranged from their establishment until April 2013. The search strategy was tailored to each database and used a combination of key terms such as “tuberculin skin test”, “TST”, “interferon-γ assay”, “quantiferon”, “ELISPOT”, “T-Spot”, “dialysis”, “tuberculosis”, and “TB”. Mesh terms (Figure S3, SDC,http://links.lww.com/TP/B67) were also applied in the search strategy. We downloaded all of the received citations into Refworks version 2.0 (RefWorks-COS, Bethesda MD, 2011). Studies were limited to the English language.
Two reviewers (B.L. and Y.X.) independently reviewed each citation by title and abstract and chose relevant articles for full-text review. Reference lists in articles selected for full-text review were further screened for studies missed by the search strategy. The full-text articles were then assessed by two reviewers (T.F. and B.L.) Articles were finalized in their inclusion in the systematic review after consultation with further reviewers (P.K. and N.T.). All disagreements were resolved by consensus.
We identified studies evaluating the diagnostic accuracy of the TST, QFT-G, and T-SPOT.TB in determining latent tuberculosis infection in patients with kidney failure on hemodialysis. The studies included had to report adequate information to ascertain and meta-analyse test sensitivity and/or specificity with regards to determining latent tuberculosis. Studies that focused on only active tuberculosis disease were excluded.
Data Extraction and Quality Assessment
A data extraction form was created to capture relevant information from the included studies. Data were independently extracted by two investigators (T.F. and B.L.) with any disagreements resolved by consensus and consultation with a third reviewer (P.K.).
We extracted the following information from each study: the type of test performed (TST, QFT-G, or TSPOT), the gold standard used to determine latent tuberculosis (contact with tuberculosis infected, history of active disease, or chest x-ray evidence of previous tuberculosis, and combinations of the aforementioned), the total size of the population tested, the number of indeterminate results of the respective test within those considered to be condition positive and negative, the total number of positive and negative cases, and the total number of positive and negative test results stratified between those who were case positive or negative. The cutoff for a positive TST was also considered in the study and the rate of BCG vaccination present in the population.
The primary outcome measure in our systematic review was the sensitivity or specificity of each test (TST, QFT-G, and T.SPOT) with regards to the diagnosis of latent tuberculosis infection in patients with kidney failure on hemodialysis.
Two reviewers (T.F. and B.L.) assessed all included studies for their quality of reporting and risk of bias using published guidelines: quality assessment of studies of diagnostic accuracy included in systematic reviews (QUADAS)34 in combination with a methodological quality evaluation applied in a previous study 27 (Fig. 4).
Data Synthesis and Analysis
Pooled sensitivity and specificity were calculated for each test using studies that were eligible for meta-analysis. The Clopper-Pearson (exact) method was used to calculate corresponding 95% confidence intervals for each individual study. Overall composite estimates for sensitivity and specificity were calculated using two methods. The first method applied a fixed effects model and pooled together true positives, true negatives, false positives, and false negatives from all studies to develop a combined estimate for sensitivity and specificity. The Clopper-Pearson method was also applied to these values to calculate a 95% confidence interval for the pooled estimates. The Cochran Q test and I-squared values were determined to assess heterogeneity for the pooled estimates. A P value less than 0.05 in the Cochran Q test implied the presence of heterogeneity among the individual study estimates. The I-squared value represents the percentage of total variation across all studies that can be attributed to heterogeneity.36 The second method made use of a random effects model to estimate pooled sensitivity and specificity while accounting for heterogeneity amongst individual studies.37 All statistical analyses were performed using SAS version 9.2 (SAS Institute, Cary, NC). Forest plots were generated using Metadata Viewer v1.04.38
1. Nuermberger E, Bishai WR, Grosset JH. Latent tuberculosis infection. Semin Respir Crit Care Med
2004; 25: 317.
2. Dheda K, Schwander SK, Zhu B, et al. The immunology of tuberculosis: from bench to bedside. Respirology
2010; 15: 433.
3. Girndt M, Sester U, Sester M, et al. Impaired cellular immune function in patients with end-stage renal failure. Nephrol Dial Transplant
1999; 14: 2807.
4. Passalent L, Khan K, Richardson R, et al. Detecting latent tuberculosis infection in hemodialysis patients: a head-to-head comparison of the T-SPOT.TB test, tuberculin skin test, and an expert physician panel. Clin J Am Soc Nephrol
2007; 2: 68.
5. Roshanravan B, Khatri M, Robinson-Cohen C, et al. A prospective study of frailty in nephrology-referred patients with CKD. Am J Kidney Dis
2012; 60: 912.
6. Richardson RM. The diagnosis of tuberculosis in dialysis patients. Semin Dial
2012; 25: 419.
7. Pai M, Riley LW, Colford JM Jr. Interferon-gamma assays in the immunodiagnosis of tuberculosis: a systematic review. Lancet Infect Dis
2004; 4: 761.
8. Hoffmann M, Tsinalis D, Vernazza P, et al. Assessment of an Interferon-gamma release assay for the diagnosis of latent tuberculosis infection in haemodialysis patient. Swiss Med Wkly
2010; 140: 286.
9. Schulz KF, Altman DG, Moher D, CONSORT Group. CONSORT 2010 statement: updated guidelines for reporting parallel group randomised trials. Int J Surg
2011; 9: 672.
10. Ates G, Yildiz T, Danis R, et al. Incidence of tuberculosis disease and latent tuberculosis infection in patients with end stage renal disease in an endemic region. Ren Fail
2010; 32: 91.
11. Chung WK, Zheng ZL, Sung JY, et al. Validity of interferon-gamma-release assays for the diagnosis of latent tuberculosis in haemodialysis patients. Clin Microbiol Infect
2010; 16: 960.
12. Habesoglu MA, Torun D, Demiroglu YZ, et al. Value of the tuberculin skin test in screening for tuberculosis in dialysis patients. Transplant Proc
2007; 39: 883.
13. Centers for Disease Control and Prevention (CDC). Tuberculosis transmission in a renal dialysis center—Nevada. MMWR Morb Mortal Wkly Rep
2003: 2004; 53: 873.
14. Poduval RD, Hammes MS. Tuberculosis screening in dialysis patients—is the tuberculin test effective? Clin Nephrol
2003; 59: 436.
15. Seyhan EC, Sokucu S, Altin S, et al. Comparison of the QuantiFERON-TB Gold In-Tube test with the tuberculin skin test for detecting latent tuberculosis infection in hemodialysis patients. Transplant Infectious Disease
2010; 12: 98.
16. Triverio P-, Bridevaux P-, Roux-Lombard P, et al. Interferon-gamma release assays versus tuberculin skin testing for detection of latent tuberculosis in chronic haemodialysis patients. Nephrol Dial Transplant
2009; 24: 1952.
17. Wauters A, Peetermans WE, Van den BP, et al. The value of tuberculin skin testing in haemodialysis patients. Nephrol Dial Transplant
2004; 19: 433.
18. Winthrop KL, Nyendak M, Calvet H, et al. Interferon-gamma release assays for diagnosing mycobacterium tuberculosis infection in renal dialysis patients. Clin J Am Soc Nephrol
2008; 3: 1357.
19. Linquist JA, Rosaia CM, Riemer B, et al. Tuberculosis exposure of patients and staff in an outpatient hemodialysis unit. Am J Infect Control
2002; 30: 307.
20. Anibarro L, Trigo M, Feijoo D, et al. Value of the tuberculin skin testing and of an interferon-gamma release assay in haemodialysis patients after exposure to M. tuberculosis. BMC Infect Dis
2012; 12: 195.
21. Ates G, Ozekinci T, Yildiz T, et al. Comparison of interferon-gamma release assay versus tuberculin skin test for latent tuberculosis screening in hemodialysis patients. Biotechnol Biotechnol Equip
2009; 23: 1242.
22. Chung WK, Zheng ZL, Kim H-, et al. Serial testing of interferon-γ-release assays for the diagnosis of latent tuberculosis in hemodialysis patients. J Infect
2010; 61: 144.
23. Inoue T, Nakamura T, Katsuma A, et al. The value of QuantiFERON TB-Gold in the diagnosis of tuberculosis among dialysis patients. Nephrol Dial Transplant
2009; 24: 2252.
24. Lee SSJ, Chou KJ, Su IJ, et al. High prevalence of latent tuberculosis infection in patients in end-stage renal disease on hemodialysis: comparison of quantiFERON-TB GOLD, ELISPOT, and tuberculin skin test. Infection
2009; 37: 96.
25. Lee SS, Chou KJ, Dou HY, et al. High prevalence of latent tuberculosis infection in dialysis patients using the interferon-gamma release assay and tuberculin skin test. Clin J Am Soc Nephrol
2010; 5: 1451.
26. Kunst H, Khan KS. Age-related risk of hepatotoxicity in the treatment of latent tuberculosis infection: a systematic review. Int J Tuberc Lung Dis
2010; 14: 1374.
27. Rogerson TE, Chen S, Kok J, et al. Tests for latent tuberculosis in people with ESRD: a systematic review. Am J Kidney Dis
2013; 61: 33.
28. Tuberculosis Prevention and Control, Public Health Agency of Canada, Canadian Lung Association. Canadian Tuberculosis Standards
. 6th ed. Canada: Minister of Health, 2007: 438.
29. Kidney Disease: Improving Global Outcomes (KDIGO) Transplant Work Group. KDIGO clinical practice guideline for the care of kidney transplant recipients. Am J Transplant
2009; 9 Suppl 3: S1.
30. Chadban SJ, Barraclough KA, Campbell SB, et al. KHA-CARI guideline: KHA-CARI adaptation of the KDIGO Clinical Practice Guideline for the Care of Kidney Transplant Recipients. Nephrology (Carlton)
2012; 17: 204.
31. Chia S, Karim M, Elwood RK, et al. Risk of tuberculosis in dialysis patients: a population-based study. Int J Tuberc Lung Dis
1998; 2: 989.
32. Bronfenbrenner M. Notes on the elasticity of derived demand. Oxford Economic Papers
1961; 13: 254.
33. Kowada A. Cost effectiveness of the interferon-gamma release assay for tuberculosis screening of hemodialysis patients. Nephrol Dial Transplant
2013; 28: 682.
34. Whiting P, Rutjes AW, Reitsma JB, et al. The development of QUADAS: a tool for the quality assessment of studies of diagnostic accuracy included in systematic reviews. BMC Med Res Methodol
2003; 3: 25.
35. World Health Organization. Global Tuberculosis Report 2012
. France: World Health Organization, 2012: 89.
36. Higgins JP, Thompson SG, Deeks JJ, et al. Measuring inconsistency in meta-analyses. BMJ
2003; 327: 557.
37. Menke J. Bivariate random-effects meta-analysis of sensitivity and specificity with SAS PROC GLIMMIX. Methods Inf Med
2010; 49: 54.
38. Boyles A, Harris S, Rooney A, et al. Forest Plot Viewer: a new graphing tool. Epidemiology
2011; 22: 746.